wtf is codex-chrome-operator?

Codex Chrome Operator is a local automation bridge that lets Codex, an AI coding tool, observe and control a Chrome browser window on a Windows machine. Instead of Codex being limited to code files, this project lets it interact with web pages: reading page content, clicking buttons, filling in forms, and taking screenshots, all while the browser window stays fully visible to the person using it. The system is made of four connected pieces: a Chrome extension, a native messaging bridge, which is a small program that lets Chrome talk to local software, a local operator daemon that acts as the central process enforcing policy, and an MCP adapter that exposes a small set of strict tools to Codex. All communication stays on the local machine, with no remote servers involved in controlling the browser. The key design principle is being conservative by default. A readiness check must pass before anything runs, requiring the daemon to be running, the extension connected, and the target site approved and not blocked by the user. High-impact actions like placing orders, submitting payments, or completing checkout are permanently blocked regardless of instructions, and cannot be unlocked through an approval prompt. An emergency stop command halts all pending actions immediately and blocks new ones until it is cleared. Low-risk actions the tool can perform include clicking, typing, scrolling, selecting options, and pressing keys. It can also read a compact snapshot of the page's structure for faster AI inspection and capture screenshots for local visual analysis. The project targets Windows with Google Chrome, requires Node.js version 24 or newer and PowerShell, and is written in JavaScript. It is aimed at developers who want an AI coding assistant to safely perform guarded, human-visible browser actions rather than run unsupervised background automation.